JP2024020237A - Predicting three-dimensional features for autonomous driving - Google Patents

Abstract

PROBLEM TO BE SOLVED: To generate accurate labeled training data for a deep learning system used in applications such as autonomous driving.

SOLUTION: Sensor data including image data and odometry data is received to create a training data set (201). A machine learning model is trained in a neural network such as a convolutional neural network (CNN) (203). Then, the trained machine learning model is installed in a vehicle (205), and sensor data is received (207). Then, the trained machine learning model by an AI processor or the like is applied (209), and the autonomous vehicle is controlled (211).

SELECTED DRAWING: Figure 2

COPYRIGHT: (C)2024,JPO&INPIT

Description

[Cross reference to related applications]
This application was filed on February 1, 2019, and is based on “PREDICTING THREE-
DIMENSIONAL FEATURES FOR AUTONOMOUS DRIV
U.S. Patent Application No. 16/2657 entitled ING (3D Feature Prediction for Autonomous Driving)
No. 20, a continuation application, which claims priority and is incorporated herein by reference in its entirety.

Deep learning systems used in applications such as autonomous driving are developed by training machine learning models. Typically, the performance of deep learning systems is limited, at least in part, by the quality of the training set used to train the model. Significant resources are often invested in collecting, curating, and annotating training data. Traditionally, much of the effort to curate training datasets is done manually by considering potential training data and appropriately labeling the data with relevant features. The effort required to create a training set with accurate labels can be significant and often tedious. Additionally, it is often difficult to collect and accurately label the data that machine learning models need to improve. Therefore, there is a need to improve the process for generating training data with accurate labeled features.

Various embodiments of the invention are disclosed in the following detailed description and accompanying drawings.

FIG. 1 is a block diagram illustrating an embodiment of a deep learning system for autonomous driving.

1 is a flow diagram illustrating one embodiment of a process for training and applying a machine learning model for autonomous driving.

2 is a flow diagram illustrating one embodiment of a process for creating training data using time series elements.

1 is a flow diagram illustrating one embodiment of a process for training and applying a machine learning model for autonomous driving.

FIG. 3 is a diagram showing an example of an image captured from a vehicle sensor.

FIG. 3 shows an example of an image captured from a vehicle sensor with a predicted three-dimensional trajectory of lane lines.

The invention relates to a process, an apparatus, a system, a composition of matter, a computer program product embodied on a computer readable storage medium, and/or a processor, e.g. A processor configured to execute instructions provided by a processor may be implemented in a variety of ways. These implementations, or any other forms the invention may take, may be referred to herein as techniques. In general, the order of steps in the disclosed processes may be varied within the scope of the invention. Unless otherwise specified, components such as processors or memory that are described as being configured to perform a task are general components that are temporarily configured to perform a task at a given time. , or may be implemented as a specific component manufactured to perform a task. As used herein, the term "processor" refers to one or more devices, circuits, and/or processing cores configured to process data, such as computer program instructions.

A detailed description of one or more embodiments of the invention is provided below, along with accompanying drawings that illustrate the principles of the invention. Although the invention is described in connection with such embodiments, the invention is not limited to any embodiment. The scope of the invention is limited only by the claims, and the invention encompasses numerous alternatives, modifications, and equivalents. Numerous specific details are set forth in the following description to provide a thorough understanding of the invention. These details are provided for purposes of illustration and the invention may be practiced according to the claims without some or all of these specific details. For clarity, technical material known in the art related to the invention has not been described in detail so as not to unnecessarily obscure the invention.

A machine learning training technique for generating highly accurate machine learning results is disclosed. A training dataset is created using data captured by sensors on the vehicle to capture the vehicle's environment and vehicle operating parameters. For example, sensors mounted on a vehicle capture data such as image data of the road on which the vehicle is traveling and the surrounding environment. Sensor data can capture vehicle lane lines, vehicle lanes, other vehicle traffic, obstacles, traffic control signs, etc. Odometric and other similar sensors are
Capture vehicle operating parameters such as vehicle speed, steering, heading, changes in direction, changes in position, changes in altitude, and changes in speed. The captured dataset is sent to a training server to create a training dataset. Training datasets are used to train machine learning models that produce highly accurate machine learning results. In some embodiments, time series acquisition data is used to generate training data. For example, ground truth (gro
and truth) is determined based on a group of time series elements and is associated with a single element from the group. As an example, a series of images over a period of time, such as 30 seconds, is used to determine the actual path of the vehicle lane line over the period of time the vehicle is traveling. Vehicle lane lines are determined by using the most accurate image of the vehicle lane over the period of time. Different parts (or positions) of the lane line can be identified from different time series of image data. As the vehicle travels along the lane line, more accurate data is captured for different parts of the lane line. In some examples, lane blockages appear when a vehicle travels, for example, along a hidden curve or over the top of a hill. The most accurate portion of lane lines from each image in the time series may be used to identify lane lines across the group of image data. Image data of distant lane lines typically has a lower level of detail than image data of lane lines near the vehicle. By capturing a time series of image data as the vehicle travels along the lane, accurate image data and corresponding odometry data of all parts of the corresponding lane line are collected.

In some embodiments, a three-dimensional representation of a feature, such as a lane line, is created from a group of time-series elements that correspond to the ground truth. This ground truth is then associated with a subset of time-series elements, such as a single image frame of a group of captured image data. For example, the first image of the image group is associated with the ground truth of lane lines represented in three-dimensional space. While the ground truth is determined based on the group of images, the selected first frame and the ground truth are used to create training data. As an example, training data is created for predicting a three-dimensional representation of a vehicle lane using only a single image. In some embodiments, any element or group of elements of the group of time series elements is associated with ground truth and used to create training data. For example, ground truth may be applied to an entire video sequence to create training data. As another example, intermediate or last elements of a group of time series elements are associated with ground truth and used to create training data.

In various embodiments, the selected image and the ground truth include lane lines,
It can be applied to various features such as route prediction of vehicles including adjacent vehicles, depth distance of objects, traffic signs, etc. For example, a series of images of a vehicle in adjacent lanes is used to predict the vehicle's path. Using a time series of images taken by neighboring vehicles and the actual route, a single image of the photographed group and the actual route can be used as training data for predicting the vehicle's path. . This information can also be used to predict whether neighboring vehicles will cut into the autonomous vehicle's path. For example, route prediction can predict whether neighboring vehicles will merge in front of the autonomous vehicle. Autonomous vehicles can be controlled to minimize the possibility of collisions. For example, autonomous vehicles can slow down to avoid collisions, adjust vehicle speed and/or steering to prevent collisions, and
or initiate a warning to the autonomous vehicle's occupants and/or change lanes, etc. In various embodiments, the ability to accurately infer route predictions, including vehicle path predictions, significantly improves the safety of autonomous vehicles.

In some embodiments, a trained machine learning model is used to predict a three-dimensional representation of one or more features for automated driving, including lane lines.
For example, instead of identifying two-dimensional lane lines from image data by segmenting an image of lane lines, a three-dimensional representation is generated using elements of a time series and odometry data corresponding to the time series. The three-dimensional representation includes elevation changes that greatly improve the accuracy of lane line detection and detection of corresponding lanes and identified drivable routes.
In some embodiments, lane lines are represented using one or more splines or another parameterized representation format. Using piecewise polynomials to represent lane lines greatly reduces the computational resources required to evaluate three-dimensional objects.
This reduction in computational resources corresponds to improved processing speed and efficiency without significantly sacrificing representation accuracy. In various embodiments, lane lines, particularly including lane line curves, may be represented using a piecewise polynomial, a three-dimensional point set, or another suitable representation. For example, piecewise polynomials interpolate actual lane lines using highly accurate sections of lane lines identified from groups of elements captured over time using sensor data.

In some embodiments sensor data is received. Sensor data may include images (such as video and/or still images), radar, audio, LiDAR, inertial, odometry, position, and/or other forms of sensor data. Sensor data includes groups of time-series elements. For example, a group of time-series elements may include a group of images captured from a vehicle's camera sensor over a period of time. In some embodiments, the training dataset determines, for at least selected time series elements within the group of time series elements, corresponding ground truth based on the plurality of time series elements within the group of time series elements. It is determined by things such as. For example, the ground truth is determined by examining the most relevant parts of each element of a group of time series elements, including previous and/or later time series elements within the group. In some scenarios, only the previous and/or later time series elements are absent from the previous time series elements, such as vehicle lane lines that first disappear around the curve and appear only in later elements of the time series. Contains data. The determined ground truth may be a three-dimensional representation of vehicle lane lines, a predicted path of the vehicle, or another similar prediction. An element of a group of time series elements is selected and associated with the ground truth. The selected elements and ground truth are part of the training dataset. In some embodiments, a processor is used to train a machine learning model using a training dataset. For example, the training dataset is used to train a machine learning model to infer features used for autonomous driving or driver assistance operations of the vehicle. Using a trained machine learning model, a neural network analyzes vehicle lanes, drivable space, objects (e.g., pedestrians, stationary vehicles, moving vehicles, etc.), weather (e.g., rain, hail, fog, etc.), traffic Features related to autonomous driving, such as control objects (e.g., traffic lights, traffic signs, road markings, etc.), traffic patterns, etc., can be inferred.

In some embodiments, a system includes a processor and memory coupled to the processor. The processor is configured to receive image data based on images captured by a camera of the vehicle. For example, a camera sensor mounted on a vehicle captures images of the vehicle's environment. The camera may be a forward-facing camera, a pillar camera, or another appropriately positioned camera. The image data captured from the camera is sent to the GPU or A
Processed using a processor such as an I processor. In some embodiments, the image data is used as the basis for input to a trained machine learning model trained to predict three-dimensional trajectories of vehicle lanes. For example, image data is used as input to a neural network trained to predict vehicle lanes. The machine learning model infers the three-dimensional trajectory of the detected lane. Instead of segmenting the image into lane and non-lane segments of a two-dimensional image, a three-dimensional representation is inferred. In some embodiments, the three-dimensional representation is a spline, a parametric curve, or another representation that can describe a curve in three dimensions. In some embodiments, a three-dimensional trajectory of the vehicle lane is provided in automatically controlling the vehicle. For example, three-dimensional trajectories are used to determine lane lines and corresponding driveable spaces.

FIG. 1 is a block diagram illustrating an embodiment of a deep learning system for autonomous driving. Deep learning systems can be used together to collect and process data for autonomous driving and/or driver assistance operations in vehicles, as well as for training machine learning models for autonomous driving. Contains various components. In various embodiments, the deep learning system is installed in a vehicle. Data from the vehicle can be used to train and improve the autonomous driving capabilities of the vehicle or other similar vehicles.

In the illustrated example, the deep learning system 100 includes a sensor 101, an image preprocessor 103
, a deep learning network 105 , an artificial intelligence (AI) processor 107 , a vehicle control module 109 , and a network interface 111 . In various embodiments, different components are communicatively connected. For example, sensor data from sensor 101 is provided to image preprocessor 103. The processed sensor data of the image preprocessor 103 is fed to a deep learning network 105 running on the AI processor 107 . The output of the deep learning network 105 running on the AI processor 107 is provided to the vehicle control module 109. In various embodiments, vehicle control module 109 is connected to and controls vehicle operations, such as vehicle speed, braking, and/or steering. In various embodiments, sensor data and/or machine learning results may be sent to a remote server via network interface 111. For example, sensor data can be sent to a remote server via network interface 111 to collect training data to improve vehicle performance, comfort, and/or safety. In various embodiments, network interface 111 communicates with a remote server, makes phone calls, among other reasons.
Used to send and/or receive text messages and transmit sensor data based on vehicle operation. In some embodiments, deep learning system 100 may include additional or fewer components as desired. For example, in some embodiments, image preprocessor 103 is an optional component. As another example, some embodiments include a post-processing component (not shown) to perform post-processing on the output of deep learning network 105 before the output is provided to vehicle control module 109. is used.

In some embodiments, sensor 101 includes one or more sensors. In various embodiments, sensors 101 may be mounted on the vehicle at different locations on the vehicle and/or oriented in one or more different directions. For example, the sensor 101 may be attached to the front, side, rear, and/or roof of the vehicle in a forward, rearward, sideways, etc. direction. In some embodiments, sensor 101 may be an image sensor, such as a high dynamic range camera. In some embodiments, sensor 101 includes a non-visual sensor. In some embodiments, sensors 101 include radar, audio, LiDAR, inertial, odometry, position, and/or ultrasound sensors, among others. In some embodiments, sensor 101 is not attached to a vehicle with vehicle control module 109. For example, sensor 101 may be attached to an adjacent vehicle, and/or attached to the road or environment, and included as part of a deep learning system to capture sensor data. In some embodiments, sensor 101 includes one or more cameras that capture the road surface on which the vehicle is traveling. For example, one or more forward-facing and/or pillar cameras capture lane markings of the lane in which the vehicle is traveling. As another example, the camera captures adjacent vehicles, including a vehicle that is attempting to cut into the lane in which the vehicle is traveling. Additional sensors capture vehicle control information including information regarding odometry, position, and/or vehicle trajectory. Sensor 101 can include both image sensors that can capture still images and/or video. Data may be captured over a period of time, such as a sequence of data captured over a period of time. For example, images of lane markings may be captured along with vehicle odometry data over a period of 15 seconds or another suitable period of time. In some embodiments, sensor 101 includes a location sensor, such as a Global Positioning System (GPS) sensor, for determining the location and/or change in location of the vehicle.

In some embodiments, image preprocessor 103 is used to preprocess sensor data of sensor 101. For example, image preprocessor 103 may be used to preprocess sensor data, divide sensor data into one or more components, and/or post-process one or more components. In some embodiments, image preprocessor 103 includes a graphics processing unit (GPU), a central processing unit (CPU),
An image signal processor or a dedicated image processor. In various embodiments, image preprocessor 103 is a tone mapper processor that processes high dynamic range data. In some embodiments, image preprocessor 103 is implemented as part of artificial intelligence (AI) processor 107. For example, image preprocessor 103 may be a component of AI processor 107. In some embodiments, image preprocessor 103 may be used to normalize images or transform images. For example, an image captured with a fisheye lens may be distorted, and image preprocessor 103 may be used to transform the image to remove or correct the distortion. In some embodiments, noise, distortion, and/or blurring is removed or reduced during a pre-processing step. In various embodiments, images are adjusted or normalized to improve the results of machine learning analysis. For example, the white balance of the image is adjusted to account for different lighting operating conditions such as daylight, clear skies, cloudy, dusk, sunrise, sunset, and nighttime conditions, among others.

In some embodiments, deep learning network 105 provides vehicle control, including analyzing the driving environment to determine lane markers, lanes, drivable spaces, obstacles, and/or potential vehicle paths, etc. A deep learning network used to determine parameters. For example, deep learning network 105 may be an artificial neural network such as a convolutional neural network (CNN) that is trained with inputs such as sensor data and whose output is provided to vehicle control module 109. As an example, the output may include at least a three-dimensional representation of the lane markers. As another example, the output may include at least potential vehicles that may merge into the lane of vehicles. In some embodiments, deep learning network 105 receives at least sensor data as input. Additional inputs may include scene data describing vehicle specifications, such as the environment surrounding the vehicle and/or operating characteristics of the vehicle. The scene data may include scene tags that describe the environment around the vehicle, such as rain, wet roads, snowfall, slush, high density traffic, highways, cities, school zones, etc. In some embodiments, the output of deep learning network 105 is a three-dimensional trajectory of the vehicle lane of the vehicle. In some embodiments, the output of deep learning network 105 is a potential vehicle interrupt. For example, deep learning network 105 identifies neighboring vehicles that are likely to enter the lane in front of the vehicle.

In some embodiments, artificial intelligence (AI) processor 107 is a hardware processor for executing deep learning network 105. In some embodiments, A
The I processor 107 applies a convolutional neural network (CNN) to the sensor data.
A dedicated AI processor for performing inference using . The AI processor 107 is
It can be optimized for the bit depth of sensor data. In some embodiments,
AI processor 107 is optimized for deep learning operations such as neural network operations including convolution, dot product, vector, and/or matrix operations, among others. In some embodiments, AI processor 107 is implemented using a graphics processing unit (GPU). In various embodiments, the AI processor 107 provides instructions to the AI processor that, when executed, cause the AI processor to perform deep learning analysis on the received input sensor data and determine machine learning results used for autonomous driving. The memory is coupled to memory configured to serve the processor. In some embodiments, AI processor 107 is used to process sensor data in preparation for making the data available as training data.

In some embodiments, vehicle control module 109 is utilized to process the output of artificial intelligence (AI) processor 107 and convert the output into vehicle control operations. In some embodiments, vehicle control module 109 is utilized to control a vehicle for autonomous driving. In various embodiments, vehicle control module 109 controls vehicle speed, acceleration,
Steering, braking, etc. can be adjusted. For example, in some embodiments, vehicle control module 109 controls the vehicle to maintain the vehicle's position within a lane, merge the vehicle into another lane, and adjust the vehicle's speed and lane placement. Used to consider merging vehicles, etc.

In some embodiments, vehicle control module 109 is used to control vehicle lighting such as brake lights, turn signals, headlights, etc. In some embodiments, vehicle control module 109 is used to control vehicle audio states, such as the vehicle's sound system, playing audio alerts, enabling microphones, and enabling horns. In some embodiments, the vehicle control module 109 controls a notification system that includes a warning system to notify the driver and/or passengers of driving events such as a potential collision or approaching an intended destination. used for. In some embodiments,
Vehicle control module 109 is used to regulate sensors such as sensor 101 in the vehicle. For example, vehicle control module 109 may be used to change orientation, output resolution, and/or
or change the parameters of one or more sensors, such as changing the format type, increasing or decreasing the capture rate, adjusting the captured dynamic range, adjusting the camera focus, enabling and/or disabling the sensor . In some embodiments, the vehicle control module 109 may be used to change parameters of the image preprocessor 103, such as changing the frequency range of the filter, adjusting feature and/or edge detection parameters, adjusting the channel and bit depth, etc. I can do it. In various embodiments, vehicle control module 109 is used to implement automated driving and/or driver assistance control of a vehicle. In some embodiments, vehicle control module 109 is implemented using a processor coupled with memory. In some embodiments, vehicle control module 109 is implemented using an application specific integrated circuit (ASIC), programmable logic device (PLD), or other suitable processing hardware.

In some embodiments, network interface 111 is a communication interface for sending and/or receiving data, including audio data. In various embodiments, network interface 111 interfaces with remote servers, connects and makes voice calls, sends and/or receives text messages, sends sensor data, and updates machine learning models. including cellular or wireless interfaces for receiving updates to deep learning networks and retrieving environmental conditions, including weather conditions and forecasts, traffic conditions, etc. For example, the network interface 111 includes the sensor 101, the image preprocessor 103, and the deep learning network 1.
05, AI processor 107, and/or vehicle control module 109 instructions and/or
or may be used to receive updates of operating parameters. The machine learning model of deep learning network 105 can be updated using network interface 111. As another example, network interface 111 may be used to update firmware of sensor 101 and/or operational parameters of image preprocessor 103, such as image processing parameters. As yet another example, network interface 111 can be used to send potential training data to a remote server for training machine learning models.

FIG. 2 is a flow diagram illustrating one embodiment of a process for training and applying machine learning models for autonomous driving. For example, input data including sensor and odometry data is received and processed to create training data for training a machine learning model. In some embodiments, the sensor data corresponds to image data captured via an autonomous driving system. In some embodiments, the sensor data corresponds to sensor data captured based on a particular use case, such as a user manually disabling autonomous driving. In some embodiments, the process is used to create and deploy machine learning models for deep learning system 100 of FIG. 1.

At 201, training data is prepared. In some embodiments, sensor data including image data and odometry data is received to create a training dataset. Sensor data may include still images and/or video from one or more cameras. Additional sensors such as radar, LiDAR, ultrasound, etc. can be used to provide relevant sensor data. In various embodiments, sensor data is paired with corresponding odometry data to help identify characteristics of the sensor data. For example, position and position change data can be used to identify the position of relevant features within sensor data such as lane lines, traffic control signals, objects, and the like. In some embodiments, the sensor data is a time series of elements and is used to determine ground truth. The group ground truth is then associated with a subset of the time series, such as the first frame of image data. The selected elements of the time series and the ground truth are used to prepare training data. In some embodiments, training data is prepared to train a machine learning model to only identify features from sensor data such as lane lines, vehicle paths, traffic patterns, etc. The prepared training data is used for training,
May contain data for validation and testing. In various embodiments, sensor data may be in different formats. For example, sensor data may be still images, videos, audio, etc. The odometry data may include vehicle operating parameters such as applied acceleration, applied braking, applied steering, vehicle position, vehicle orientation, change in vehicle position, change in vehicle orientation. In various embodiments, training data is curated and annotated to create a training dataset. In some embodiments, part of the training data preparation may be performed by a human curator. In various embodiments, some of the training data is automatically generated from data captured from the vehicle, greatly reducing the effort and time required to build a robust training data set. In some embodiments, the format of the data is compatible with machine learning models used in deployed deep learning applications. In various embodiments, the training data includes validation data for testing the accuracy of the trained model.

At 203, a machine learning model is trained. For example, a machine learning model is trained using the data prepared in 201. In some embodiments, the model is a neural network, such as a convolutional neural network (CNN). In various embodiments, the model includes multiple intermediate layers. In some embodiments, a neural network may include multiple layers, including multiple convolutional layers and pooling layers. In some embodiments, the training model is validated using a validation dataset created from received sensor data. In some embodiments, a machine learning model is trained to predict three-dimensional representations of features from a single input image. For example, a three-dimensional representation of lane lines can be inferred from images captured from a camera. As another example, an image captured from a camera predicts the predicted path of an adjacent vehicle, including whether the vehicle is about to merge.

At 205, the trained machine learning model is deployed. For example, the trained machine learning model is installed in a vehicle as an update to a deep learning network, such as deep learning network 105 of FIG. In some embodiments, over-the-air updates are used to install newly trained machine learning models. In some embodiments, the update is a firmware update sent using a wireless network, such as WiFi or a cellular network. In some embodiments, a new machine learning model may be installed when the vehicle is serviced.

At 207, sensor data is received. For example, sensor data is captured from one or more sensors of a vehicle. In some embodiments, the sensor is sensor 10 of FIG.
It is 1. The sensors can include image sensors such as a fisheye camera mounted behind the windshield, a front-facing or side-facing camera mounted on a pillar, a rear-facing camera, and the like. In various embodiments, the sensor data is in or converted to a format for use as input by the trained machine learning model at 203. For example, sensor data is
It may be raw or processed image data. In some embodiments, the data is
Data captured from ultrasound sensors, radar, LiDAR sensors, microphones, or other suitable technology. In some embodiments, the sensor data is preprocessed using an image preprocessor, such as image preprocessor 103 of FIG. 1, during a preprocessing step.
For example, images may be normalized to remove distortion, noise, etc.

At 209, the trained machine learning model is applied. For example, the machine learning model trained at 203 is applied to the sensor data received at 207. In some embodiments, application of the model is performed by an AI processor, such as AI processor 107 of FIG. 1, using a deep learning network, such as deep learning network 105 of FIG.
In various embodiments, three-dimensional representations of features, such as lane lines, are identified and/or predicted by applying a trained machine learning model. For example, two splines are inferred that represent the lane lines of the lane in which the vehicle is traveling. As another example, a predicted path for an adjacent vehicle is inferred, including whether the adjacent vehicle is likely to cut into the current lane. In various embodiments, machine learning models are applied to identify vehicles, obstacles, lanes, traffic control signals, map features, object distances, speed limits, drivable spaces, and the like.
In some embodiments, features are identified in three dimensions.

At 211, the autonomous vehicle is controlled. For example, one or more automated driving functions may be implemented by controlling various aspects of the vehicle. Examples include controlling the steering, speed, acceleration, and/or braking of a vehicle; maintaining the vehicle's position within a lane; maintaining the vehicle's position relative to other vehicles and/or obstacles; This may include providing notices or warnings. Based on the analysis performed at 209, the steering and speed of the vehicle are controlled to maintain the vehicle between the two lane lines. For example, left and right lane lines are predicted and corresponding vehicle lanes and drivable spaces are identified. In various embodiments, a vehicle control module, such as vehicle control module 109 of FIG. 1, controls the vehicle.

FIG. 3 is a flow diagram illustrating one embodiment of a process for creating training data using time-series elements. For example, a time series of elements consisting of sensor and odometry data is collected from a vehicle and used to automatically create training data. In various embodiments, the process of FIG. 3 is used to automatically label training data with the corresponding ground truth. Results corresponding to a time series are associated with elements of the time series. The results and selected elements are packaged as training data to predict future results. In various embodiments, sensors and associated data are captured using the deep learning system of FIG. For example, in various embodiments, sensor data may include sensor 1 of FIG.
Captured from 01. In some embodiments, the process of FIG. 3 is performed at 201 of FIG. In some embodiments, the process of FIG. 3 is performed to automatically collect data when existing predictions are incorrect or can be improved. For example, predictions are made by the autonomous vehicle to determine whether the vehicle is cutting into the autonomous vehicle's path. After waiting for a period of time and analyzing the captured sensor data, a determination can be made whether the prediction is correct or incorrect. In some embodiments, a determination is made that the prediction can be improved. If the prediction was incorrect or could be improved, apply the process in Figure 3 to the data associated with the prediction to create a curated set of examples to improve the machine learning model. I can do it.

At 301, a time series of elements is received. In various embodiments, the element is sensor data, such as image data captured on a vehicle and sent to a training server. Sensor data is captured over a period of time to create a time series element. In various embodiments, the elements are timestamps to maintain order of the elements. As elements progress through the time series, later events in the time series are used to help predict outcomes from earlier elements in the time series. For example, the time series may capture vehicles in adjacent lanes signaling to merge, accelerating, and positioning themselves near nearby lane lines. Using the entire time series, the results can be used to determine when a vehicle merges into a shared lane. This result can be used to predict vehicle merging based on selected elements of the time series, such as one of the initial images of the time series. As another example, the time series captures the curve of lane lines. The time series captures various dips, bends, peaks, etc. in the lane that are not apparent from a single element of the time series alone. In various embodiments, the element is sensor data in a form that the machine learning model uses as input. For example, sensor data may be raw or processed image data. In some embodiments, the data is data captured from an ultrasound sensor, radar, LiDAR sensor, or other suitable technology.

In various embodiments, a time series is organized by associating a timestamp with each element of the time series. For example, a timestamp is associated with at least a first element in the time series. Timestamps may be used to calibrate time series elements with related data such as odometry data. In various embodiments, the length of the time series is 10 seconds, 30
It may be a fixed length of time, such as seconds, or another suitable length. The length of time may be configurable. In various embodiments, the time series can be based on the speed of the vehicle, such as the average speed of the vehicle. For example, at a slower speed, the time length of the time series can be increased to capture data over a greater distance traveled than would be possible using a shorter time length for the same speed. In some embodiments, the number of elements within the time series is configurable. For example, the number of elements can be based on the distance traveled. For example, for a certain period of time,
Faster moving vehicles include more elements in the time series than slower moving vehicles. Additional elements can increase the fidelity of the captured environment and improve the accuracy of predicted machine learning results. In various embodiments, the number of elements is adjusted by adjusting the frames per second that the sensor captures data and/or by discarding unnecessary intermediate frames.

At 303, data related to elements of the time series is received. In various embodiments,
Related data is received at the training server along with the elements received at 301. In some embodiments, the relevant data is vehicle odometry data. Position, orientation, change in position, change in orientation, and/or other relevant vehicle data may be used to label the position data of the identified feature in the time series element. For example, by examining the time series of lane line elements, lane lines can be labeled with very precise locations. Typically, the lane line closest to the vehicle camera is accurate and closely related to the vehicle's location. On the other hand, the XYZ position of the line farthest from the vehicle is difficult to determine. Distant portions of the lane line may be occluded (e.g., behind a turn or hill) and/or difficult to capture accurately (e.g., due to distance, lighting, etc.). The data related to the elements is used to label parts of the identified features in the identified time series with high accuracy. In various embodiments, a threshold is used to determine whether to associate an identified portion of a feature (such as a portion of a lane line) with related data. For example, parts of the lane line that are identified with high accuracy (e.g., parts closer to the vehicle) are associated with relevant data, and parts of the lane line that are identified with less than a threshold (e.g., parts far from the vehicle)
is not associated with the element's associated data. Instead, another element of the time series and its associated data, such as a successor element with higher accuracy, is used. In some embodiments,
The relevant data is the output of a neural network, such as the output of deep learning network 105 in FIG. In some embodiments, the related data may include vehicle control module 10 of FIG.
This is the output of a vehicle control module such as 9. Related data includes speed, change in speed, acceleration,
Vehicle operating parameters such as acceleration changes, steering, steering changes, braking, braking changes, etc. may be included. In some embodiments, the relevant data is radar data for estimating the distance of objects such as obstacles.

In some embodiments, the data associated with the time series elements includes map data. For example, at 303, offline data such as road and/or satellite level map data is received. Map data can be used to identify features such as roads, vehicle lanes, intersections, speed limits, school zones, etc. For example, the map data may describe the route of vehicle lanes. As another example, the map data may describe speed limits associated with various roads in the map.

In various embodiments, data related to time-series elements is organized by associating time stamps with related data. The two datasets can be synchronized using time series elements and corresponding timestamps from related data. In some embodiments, data is synchronized at the time of acquisition. For example, once each element of the time series is captured,
A corresponding set of related data is captured and stored with the time series elements. In various embodiments, the period of related data is configurable and/or matches the period of the element's chronology. In some embodiments, the related data is sampled at the same rate as the time series elements.

At 305, ground truth is determined for the time series. In various embodiments, the time series includes machine learning features.
e) is analyzed to determine the ground truth associated with For example, a lane line is identified from a time series that corresponds to the ground truth of that lane line. As another example, the ground truth for the path of a moving object (eg, vehicle, pedestrian, bicycle, animal, etc.) is the path identified for the moving object detected from the time series. In some embodiments, a moving vehicle is annotated as an intruding vehicle if the moving vehicle enters the autonomous vehicle's lane over time. In some embodiments, the ground truth is represented as a three-dimensional representation, such as a three-dimensional trajectory. For example, the ground truth associated with lane lines may be represented as a three-dimensional parameterized spline or curve. As another example, the predicted path of the detected vehicle is determined and represented as a three-dimensional trajectory. The predicted route may be used to determine whether a vehicle is merging into an occupied space. In various embodiments, only by examining elements of the time series,
The ground truth can be determined. For example, analyzing only a subset of the time series may leave some lane lines occluded. By extending the analysis across time-series elements, lane line blockages are revealed. Furthermore, data captured near the end of the time series more accurately (eg, with higher fidelity) captures details of portions of the lane line that are more distant in distance. Furthermore, because the relevant data is based on data captured more closely (both in distance and time), the relevant data is also more accurate. In various embodiments, different parts of the detected object, such as lane lines identified by different elements of the time series of elements, can be simultaneously Location and mapping techniques are applied. The set of mapped three-dimensional locations represents the ground truth of an object, such as a segment of a lane line, captured over time. In some embodiments, the localization and mapping technique provides a precise point set, eg, a point set corresponding to different points along a vehicle lane line. Point sets can be converted to more efficient formats such as spline curves or parametric curves. In some embodiments, ground truth is determined to detect objects such as lane lines, drivable spaces, traffic control means, vehicles, etc. in three dimensions.

In some embodiments, ground truth is determined to predict semantic labels. For example, a detected vehicle may be labeled as being in the left or right lane. In some embodiments, a detected vehicle may be labeled as being in a blind spot, as a vehicle to yield to, or with another suitable semantic label. In some embodiments, vehicles are assigned to roads or lanes in the map based on the determined ground truth. As an additional example, the determined ground truth can be used to label traffic lights, lanes, drivable spaces, or other features that aid autonomous driving.

In some embodiments, the relevant data is depth (or distance) data of the detected object. By associating distance data with objects identified in a time series of elements,
A machine learning model can be trained to estimate object distance by using relevant distance data as ground truth about detected objects. In some embodiments, the distance is the distance of a detected object, such as an obstacle, barrier, moving vehicle, stationary vehicle, traffic control signal, pedestrian, etc.

At 307, the training data is packaged. For example, elements of the time series are selected and associated with the determined ground truth at 305. In various embodiments, the selected element is an initial element in the time series. The selected elements represent the sensor data input to the machine learning model, and the ground truth represents the predicted outcome. In various embodiments, training data is packaged and prepared as training data. In some embodiments, training data is packaged into training, validation, and test data. Based on the determined ground truth and selected time series elements, the training data can be used to analyze lane lines, vehicle predicted path, speed limits, vehicle cut-ins, objects, among other useful features for automated driving. It can be packaged to train machine learning models to identify distances and/or drivable spaces. The packaged training data is then available for training machine learning models.

FIG. 4 is a flow diagram illustrating one embodiment of a process for training and applying machine learning models for autonomous driving. In some embodiments, the process of FIG. 4 is utilized to collect and maintain sensor data and odometry data for training machine learning models for autonomous driving. In some embodiments, the process of FIG. 4 is performed in an autonomously enabled vehicle, whether or not autonomous driving controls are enabled. For example, sensor and odometry data may be collected immediately after autonomous driving is disengaged, while the vehicle is being operated by a human driver, and/or while the vehicle is being autonomously operated. In some embodiments, the technique illustrated by FIG. 4 is implemented using the deep learning system of FIG. 1. In some embodiments, a portion of the process of FIG. 4 may include 207, 209, and/or 2 of FIG.
11.

At 401, sensor data is received. For example, a vehicle equipped with sensors captures sensor data and provides the sensor data to a neural network operating on the vehicle.
In some embodiments, the sensor data may be visual data, ultrasound data, LiDAR data, or other suitable sensor data. For example, images are captured from a high dynamic range forward facing camera. As another example, ultrasound data is captured from a lateral ultrasound sensor. In some embodiments, a vehicle is equipped with multiple sensors to capture data. For example, in some embodiments, eight surround cameras are mounted on the vehicle to provide a 360 degree view of the vehicle's surroundings with a reach of up to 250 meters. In some embodiments, the camera sensors include a wide-angle front camera, a narrow-angle front camera, a rear-viewing camera, a front-looking side camera, and/or a rear-viewing side camera. In some embodiments, ultrasound and/or radar sensors are used to capture surrounding details. For example, 12 ultrasonic sensors can be installed in a vehicle to detect both hard and soft objects. In some embodiments, forward-facing radar is utilized to capture data of the surrounding environment. In various embodiments, radar sensors can capture details of the surroundings despite heavy rain, fog, dust, and other vehicles. Various sensors are used to capture the environment around the vehicle, and the captured data is provided for deep learning analysis.

In some embodiments, the sensor data includes vehicle position, orientation, change in position, and/or
or include odometry data, including changes in orientation, etc. For example, location data is captured and
Associated with other sensor data captured during the same time frame. As an example, location data captured during imaging of the image data is used to associate location information with the image data.

At 403, the sensor data is preprocessed. In some embodiments, one or more pre-processing passes may be performed on the sensor data. For example, the data may be pre-processed to remove noise, correct for alignment issues and/or blurring, etc. In some embodiments, one or more different filtering passes are performed on the data. For example, a high pass filter can be performed on the data and a low pass filter can be performed on the data to separate different components of the sensor data. In various embodiments, the preprocessing steps performed at 403 are optional and/or may be incorporated into the neural network.

At 405, deep learning analysis of the sensor data is initiated. In some embodiments, deep learning analysis is performed on the optionally preprocessed sensor data at 403. In various embodiments, deep learning analysis is performed using a convolutional neural network (CNN).
It is performed using neural networks such as In various embodiments, a machine learning model is trained offline using the process of FIG. 2 and deployed to a vehicle to perform inference on sensor data. For example, a model may be trained to properly identify road lane lines, obstacles, pedestrians, moving vehicles, parked vehicles, drivable spaces, etc.
In some embodiments, multiple trajectories of lane lines are identified. For example, several potential trajectories of lane lines are detected, each trajectory having a corresponding probability of occurrence.
In some embodiments, the predicted lane line is the lane line with the highest probability of occurrence and/or the highest associated confidence value. In some embodiments, predicted lane lines from deep learning analysis are required to exceed a minimum confidence threshold. In various embodiments, the neural network includes multiple layers including one or more hidden layers.
In various embodiments, sensor data and/or deep learning analysis results are retained and transmitted at 411 for automatic generation of training data.

In various embodiments, deep learning analysis is used to predict additional features.
Predicted features can be used to assist autonomous driving. For example, detected vehicles can be assigned to lanes or roads. As another example, the detected vehicle may be in a blind spot, should yield, be in the lane to the left, be in the lane to the right, or have another appropriate attribute. can be determined. Similarly, deep learning analysis can identify traffic lights, drivable spaces, pedestrians, obstacles, or other suitable features for driving.

At 407, the results of the deep learning analysis are provided to the vehicle control. For example, the results are provided to a vehicle control module to control the vehicle for autonomous driving and/or to perform autonomous driving functions. In some embodiments, the results of the deep learning analysis at 405 are passed through one or more additional deep learning passes using one or more different machine learning models. For example, a predicted path of lane lines can be used to determine a vehicle lane, and the determined vehicle lane is used to determine a drivable space. The drivable space is then used to route the vehicle. Similarly, in some embodiments, anticipated vehicle interruptions are detected. The vehicle's determined route takes into account predicted interruptions to avoid potential collisions. In some embodiments, the various outputs of deep learning are
It is used to build a three-dimensional representation of the vehicle's environment for autonomous driving, including the vehicle's predicted route, identified obstacles, identified traffic control signals including speed limits, etc. In some embodiments, the vehicle control module utilizes the determined results to control the vehicle along the determined route. In some embodiments, the vehicle control module is vehicle control module 109 of FIG.

At 409, the vehicle is controlled. In some embodiments, an autonomously activated vehicle is controlled using a vehicle control module, such as vehicle control module 109 of FIG. The vehicle control may, for example, adjust the speed and/or steering of the vehicle to maintain the vehicle in the lane at an appropriate speed considering the surrounding environment. In some embodiments, the results are used to adjust the vehicle in anticipation of adjacent vehicles merging into the same lane. In various embodiments, using the results of the deep learning analysis, the vehicle control module:
Determining an appropriate way to move the vehicle, for example along the determined route at an appropriate speed.
In various embodiments, the results of vehicle controls such as speed changes, brake applications, steering adjustments, etc. are retained and used for automatic generation of training data. In various embodiments, vehicle control parameters are retained and transmitted at 411 for automatic generation of training data.

At 411, sensor data and related data are transmitted. For example, the sensor data received at 401, along with the results of the deep learning analysis at 405 and/or the vehicle control parameters used at 409, are sent to a computer server for automatic generation of training data. In some embodiments, the data is time series data and the various collected data are correlated together by a computer server. For example, odometry data is associated with captured image data to generate ground truth. In various embodiments, the collected data is transmitted wirelessly from the vehicle to the training data center, such as via WiFi or a cellular connection. In some embodiments, metadata is sent along with the sensor data. For example, metadata may include time, timestamp, location, vehicle type, and vehicle control and/or operating parameters such as speed, acceleration, braking, whether autonomous driving is enabled, steering angle, odometry data, etc. I can do it. Additional metadata includes the time since the last previous sensor data was sent, vehicle type, weather conditions, road conditions, etc. In some embodiments, the transmitted data is anonymized, for example by removing the vehicle's unique identifier. As another example, data from similar vehicle models may be merged such that individual users and their use of the vehicle are not identified.

In some embodiments, data is sent only in response to a trigger. For example, in some embodiments, incorrect predictions can be used to automatically collect sensor data and related data to create a curated set of examples to improve the predictions of a deep learning network. Trigger the transmission of. For example, the prediction performed at 405 regarding whether a vehicle is about to merge is determined to be incorrect by comparing the prediction to the observed actual result. Data, including sensor data and related data associated with the incorrect predictions, is then transmitted and used to automatically generate training data.
In some embodiments, triggers are used to detect sharp turns, road forks, lane merging, sudden stops, or other suitable scenarios where additional training data would be useful and may be difficult to collect. Specific scenarios can be identified. For example, the trigger can be based on a sudden stop or release of an autonomous driving function. As another example, vehicle operating characteristics such as changes in speed or changes in acceleration may form the basis of the trigger. In some embodiments, predictions with an accuracy below a certain threshold trigger transmission of sensor data and related data. For example, in certain scenarios, a prediction may not have a Boolean true or false outcome and is instead evaluated by determining an accuracy value for the prediction.

In various embodiments, the sensor data and related data are captured over a period of time;
The entire time series data is sent together. The time period may be configured and/or based on one or more factors such as vehicle speed, distance traveled, change in speed, etc. In some embodiments, the sampling rate of captured sensor data and/or related data is configurable. For example, the sampling rate may be increased at high speeds, during hard braking, hard acceleration, hard steering, or other suitable scenarios where additional fidelity is required.

FIG. 5 is a diagram illustrating an example of an image captured from a vehicle sensor. In the illustrated example, the image of FIG. 5 is image data 50 captured from a vehicle traveling in a lane between two lane lines.
Contains 0. The vehicle and sensor locations used to capture image data 500 are represented by label A. Image data 500 is sensor data and may be captured from a camera sensor, such as a front-facing camera of a vehicle, while driving. Image data 500 captures portions of lane lines 501 and 511. Lane lines 501 and 511 curve to the right as lane lines 501 and 511 approach the horizon. In the illustrated example, lane lines 501 and 511 are visible, but as they curve further away from the camera sensor location they become increasingly difficult to detect. The white lines drawn above the lane lines 501 and 511 are drawn from the image data 500 without any additional input.
11 detectable parts are approximated. In some embodiments, detected portions of lane lines 501 and 511 may be detected by segmenting image data 500.

In some embodiments, labels A, B, and C correspond to different locations on the road and different times in the time series. Label A corresponds to the time and location of the vehicle when image data 500 was captured. Label B is a position on the road ahead of the position of label A,
It corresponds to the position at a time after the time of label A. Similarly, label C is a position on the road in front of the position of label B, and corresponds to a position at a time after the time of label B.
As the vehicle travels, it passes the locations of label A, label B, and label C (from label A to label C) and captures a time series of sensor data and related data while traveling. The time series includes elements captured at positions (and times) with labels A, B, and C. Label A corresponds to the first element of the time series, label B corresponds to the middle element of the time series, and label C
corresponds to the middle (or potentially last) element of the time series. At each label, additional data is captured, such as odometry data for the vehicle at the label location. Depending on the length of the time series, additional or less data is captured. In some embodiments, a timestamp is associated with each element of the time series.

In some embodiments, the ground truth of lane lines 501 and 511 (
(not shown) is determined. For example, using the processes disclosed herein, the locations of lane lines 501 and 511 are identified by identifying different portions of lane lines 501 and 511 from different elements of the time series. In the illustrated example, portions 503 and 513 are identified using image data 500 and associated data (such as odometry data) acquired at the location and time of label A. Portions 505 and 515
is the image data (not shown) and related data (not shown) acquired at the position and time of label B.
odometry data). Portions 507 and 517 are identified using image data (not shown) and associated data (such as odometry data) acquired at the location and time of label C. By analyzing time series elements, lane line 50
The locations of different parts of 1 and 511 can be identified and the ground truth can be determined by combining the different identified parts. In some embodiments, the portions are identified as points along each portion of the lane line. In the illustrated example, only three portions of each lane line are highlighted to illustrate the process (portions 503, 505, and 507 of lane line 501 and portions 513, 515, and 517 of lane line 511). However, additional portions may be captured over time to determine the lane line position with higher resolution and/or greater accuracy.

In various embodiments, the location of the portion of the image data capturing lane lines 501 and 511 that is closest to the location of the sensor is determined with high accuracy. For example, portions 503 and 5
Position 13 is the image data 500 of label A and related data (odometry data, etc.)
identified with high accuracy using The locations of portions 505 and 515 are identified with high accuracy using the label B image data and related data. The positions of portions 507 and 517 are
It is identified with high accuracy using the image data of label C and related data. By utilizing elements of the time series, the positions of various parts of the lane lines 501 and 511 captured by the time series are identified with high precision in three dimensions and used as the basis for the ground truth of the lane lines 501 and 511. can do. In various embodiments, the determined ground truth is associated with selected elements of a time series, such as image data 500. The ground truth and selected elements may be used to create training data for predicting lane lines. In some embodiments, training data is created automatically without human labeling. The training data can be used to train a machine learning model to predict three-dimensional trajectories of lane lines from captured image data, such as image data 500.

FIG. 6 is a diagram illustrating an example of an image captured from a vehicle sensor with a predicted three-dimensional trajectory of lane lines. In the illustrated example, the image of FIG. 6 includes image data 600 captured from a vehicle traveling in a lane between two lane lines. The location of the vehicle and sensor used to capture image data 600 is represented by label A. In some embodiments, label A corresponds to the same location as label A in FIG. Image data 600 is sensor data and may be captured from a camera sensor, such as a front-facing camera of a vehicle, while driving. Image data 600 captures portions of lane lines 601 and 611. lane line 6
01 and 611 curve to the right as lane lines 601 and 611 approach the horizon. In the illustrated example, lane lines 601 and 611 are visible, but become increasingly difficult to detect as they curve further away from the camera sensor location.
The red lines drawn above lane lines 601 and 611 are
11 predicted 3D trajectories. Using the process disclosed herein, a three-dimensional trajectory is predicted using image data 600 as input to a trained machine learning model. In some embodiments, the predicted three-dimensional trajectory is represented as a three-dimensional parameterized spline or another parameterized representation format.

In the illustrated example, portion 621 of lane lines 601 and 611 is a portion of lane lines 601 and 611 that are separated by a distance. Part 6 of lane lines 601 and 611
The three-dimensional position (i.e., longitude, latitude, and altitude) of 21 is determined with high accuracy using the process disclosed herein and included in the predicted three-dimensional trajectory of lane lines 601 and 611. Using a trained machine learning model, using image data 600 and without requiring position data at the location of portion 621 of lane lines 601 and 611,
The three-dimensional trajectory of lane lines 601 and 611 can be predicted. In the illustrated example,
Image data 600 is captured at the location and time of label A.

In some embodiments, label A in FIG. 6 corresponds to label A in FIG.
The predicted three-dimensional trajectories of 01 and 611 are determined using only image data 600 as input to the trained machine learning model. By training a machine learning model using the ground truth determined using the time series of image data and associated data containing the elements acquired at the locations of labels A, B, and C in Figure 5, the lane lines 601 and 61
1's three-dimensional trajectory is predicted with high accuracy even for distant lane line sections, such as section 621. Although image data 600 and image data 500 of FIG. 5 are related, trajectory prediction does not require image data 600 to be included in the training data. By training with enough training data, it can predict lane lines even in newly encountered scenarios. In various embodiments, the predicted three-dimensional trajectory of lane lines 601 and 611 may be used to maintain the position of the vehicle within the detected lane line and/or to move the vehicle along the detected lane of the predicted lane line. Used to navigate autonomously. Predicting lane lines in three dimensions greatly improves navigation performance, safety, and accuracy.

Although the foregoing embodiments have been described in some detail for clarity of understanding, the invention is not limited to the details provided. There are many alternative ways of implementing the invention. The disclosed embodiments are illustrative and not restrictive.

Claims (20)

A processor,
receiving image data based on images captured by a camera of the vehicle;
using the image data as a basis for input to a trained machine learning model trained to predict three-dimensional trajectories of machine learning features;
a processor configured to provide the three-dimensional trajectory of the machine learning feature when automatically controlling the vehicle;
a memory coupled to the processor and configured to provide instructions to the processor. The system of claim 1, wherein the machine learning feature is associated with vehicle lane lines. 3. The system of claim 2, wherein a portion of the three-dimensional trajectory of the vehicle lane line is occluded in the image data. 4. The system of claim 2 or 3, wherein the processor is further configured to use the vehicle lane lines to identify vehicle lanes occupied by the vehicle. 5. A system according to any one of claims 2 to 4, wherein vehicle lanes are used to identify drivable spaces. 6. The system of claim 5, wherein the identified drivable space is used to maintain the vehicle within the vehicle lane. 7. A system according to any preceding claim, wherein the camera of the vehicle comprises at least one of a forward-facing camera, a side-facing camera, or a rear-facing camera. 8. The system of any preceding claim, wherein the predicted three-dimensional trajectory is one of a plurality of identified potential trajectories. 9. The system of claim 8, wherein the predicted three-dimensional trajectory has the highest probability of occurrence value among the plurality of identified potential trajectories. 10. The system of any one of claims 1 to 9, wherein training data used to train the trained machine learning model comprises selected time series elements within a group of time series elements. 11. The system of claim 10, wherein the training data further includes an associated set of odometry data associated with the group of time series elements. 12. The system of claim 10 or 11, wherein at least part of the training set is automatically labeled with a corresponding ground truth. Claims 1-1, wherein controlling the vehicle includes adjusting speed and steering of the vehicle.
2. The system according to any one of 2. 14. A system according to any one of claims 1 to 13, wherein the three-dimensional trajectory is represented by a spline. 15. A system according to any preceding claim, wherein the three-dimensional trajectory is represented by one or more piecewise polynomials. 16. A system according to any preceding claim, wherein the machine learning feature is associated with a predicted route of a second vehicle. The predicted route of the second vehicle is used to determine whether the second vehicle is likely to enter a lane occupied by the automatically controlled vehicle. The system according to item 16. 18. The system of claim 17, wherein controlling the vehicle includes adjusting speed or steering to avoid collision with the second vehicle. A computer program product, the computer program product being embodied in a non-transitory computer readable storage medium; and
receiving image data based on images captured by a camera of the vehicle;
using the image data as a basis for input to a trained machine learning model trained to predict three-dimensional trajectories of vehicle lanes;
A computer program product comprising computer instructions for providing the three-dimensional trajectory of the vehicle lane when automatically controlling the vehicle. receiving image data based on images captured by a camera of the vehicle;
using the image data as a basis for input to a trained machine learning model trained to predict three-dimensional trajectories of vehicle lanes;
providing the three-dimensional trajectory of the vehicle lane when automatically controlling the vehicle.

Images